datafusion_datasource_csv/
source.rs

1// Licensed to the Apache Software Foundation (ASF) under one
2// or more contributor license agreements.  See the NOTICE file
3// distributed with this work for additional information
4// regarding copyright ownership.  The ASF licenses this file
5// to you under the Apache License, Version 2.0 (the
6// "License"); you may not use this file except in compliance
7// with the License.  You may obtain a copy of the License at
8//
9//   http://www.apache.org/licenses/LICENSE-2.0
10//
11// Unless required by applicable law or agreed to in writing,
12// software distributed under the License is distributed on an
13// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
14// KIND, either express or implied.  See the License for the
15// specific language governing permissions and limitations
16// under the License.
17
18//! Execution plan for reading CSV files
19
20use datafusion_datasource::schema_adapter::SchemaAdapterFactory;
21use std::any::Any;
22use std::fmt;
23use std::io::{Read, Seek, SeekFrom};
24use std::sync::Arc;
25use std::task::Poll;
26
27use datafusion_datasource::decoder::{deserialize_stream, DecoderDeserializer};
28use datafusion_datasource::file_compression_type::FileCompressionType;
29use datafusion_datasource::file_meta::FileMeta;
30use datafusion_datasource::file_stream::{FileOpenFuture, FileOpener};
31use datafusion_datasource::{
32    as_file_source, calculate_range, FileRange, ListingTableUrl, RangeCalculation,
33};
34
35use arrow::csv;
36use arrow::datatypes::SchemaRef;
37use datafusion_common::{DataFusionError, Result, Statistics};
38use datafusion_common_runtime::JoinSet;
39use datafusion_datasource::file::FileSource;
40use datafusion_datasource::file_scan_config::FileScanConfig;
41use datafusion_execution::TaskContext;
42use datafusion_physical_plan::metrics::ExecutionPlanMetricsSet;
43use datafusion_physical_plan::{
44    DisplayFormatType, ExecutionPlan, ExecutionPlanProperties,
45};
46
47use crate::file_format::CsvDecoder;
48use futures::{StreamExt, TryStreamExt};
49use object_store::buffered::BufWriter;
50use object_store::{GetOptions, GetResultPayload, ObjectStore};
51use tokio::io::AsyncWriteExt;
52
53/// A Config for [`CsvOpener`]
54///
55/// # Example: create a `DataSourceExec` for CSV
56/// ```
57/// # use std::sync::Arc;
58/// # use arrow::datatypes::Schema;
59/// # use datafusion_datasource::file_scan_config::{FileScanConfig, FileScanConfigBuilder};
60/// # use datafusion_datasource::PartitionedFile;
61/// # use datafusion_datasource_csv::source::CsvSource;
62/// # use datafusion_execution::object_store::ObjectStoreUrl;
63/// # use datafusion_datasource::source::DataSourceExec;
64///
65/// # let object_store_url = ObjectStoreUrl::local_filesystem();
66/// # let file_schema = Arc::new(Schema::empty());
67///
68/// let source = Arc::new(CsvSource::new(
69///         true,
70///         b',',
71///         b'"',
72///     )
73///     .with_terminator(Some(b'#')
74/// ));
75/// // Create a DataSourceExec for reading the first 100MB of `file1.csv`
76/// let config = FileScanConfigBuilder::new(object_store_url, file_schema, source)
77///     .with_file(PartitionedFile::new("file1.csv", 100*1024*1024))
78///     .with_newlines_in_values(true) // The file contains newlines in values;
79///     .build();
80/// let exec = (DataSourceExec::from_data_source(config));
81/// ```
82#[derive(Debug, Clone, Default)]
83pub struct CsvSource {
84    batch_size: Option<usize>,
85    file_schema: Option<SchemaRef>,
86    file_projection: Option<Vec<usize>>,
87    pub(crate) has_header: bool,
88    delimiter: u8,
89    quote: u8,
90    terminator: Option<u8>,
91    escape: Option<u8>,
92    comment: Option<u8>,
93    metrics: ExecutionPlanMetricsSet,
94    projected_statistics: Option<Statistics>,
95    schema_adapter_factory: Option<Arc<dyn SchemaAdapterFactory>>,
96}
97
98impl CsvSource {
99    /// Returns a [`CsvSource`]
100    pub fn new(has_header: bool, delimiter: u8, quote: u8) -> Self {
101        Self {
102            has_header,
103            delimiter,
104            quote,
105            ..Self::default()
106        }
107    }
108
109    /// true if the first line of each file is a header
110    pub fn has_header(&self) -> bool {
111        self.has_header
112    }
113    /// A column delimiter
114    pub fn delimiter(&self) -> u8 {
115        self.delimiter
116    }
117
118    /// The quote character
119    pub fn quote(&self) -> u8 {
120        self.quote
121    }
122
123    /// The line terminator
124    pub fn terminator(&self) -> Option<u8> {
125        self.terminator
126    }
127
128    /// Lines beginning with this byte are ignored.
129    pub fn comment(&self) -> Option<u8> {
130        self.comment
131    }
132
133    /// The escape character
134    pub fn escape(&self) -> Option<u8> {
135        self.escape
136    }
137
138    /// Initialize a CsvSource with escape
139    pub fn with_escape(&self, escape: Option<u8>) -> Self {
140        let mut conf = self.clone();
141        conf.escape = escape;
142        conf
143    }
144
145    /// Initialize a CsvSource with terminator
146    pub fn with_terminator(&self, terminator: Option<u8>) -> Self {
147        let mut conf = self.clone();
148        conf.terminator = terminator;
149        conf
150    }
151
152    /// Initialize a CsvSource with comment
153    pub fn with_comment(&self, comment: Option<u8>) -> Self {
154        let mut conf = self.clone();
155        conf.comment = comment;
156        conf
157    }
158}
159
160impl CsvSource {
161    fn open<R: Read>(&self, reader: R) -> Result<csv::Reader<R>> {
162        Ok(self.builder().build(reader)?)
163    }
164
165    fn builder(&self) -> csv::ReaderBuilder {
166        let mut builder = csv::ReaderBuilder::new(Arc::clone(
167            self.file_schema
168                .as_ref()
169                .expect("Schema must be set before initializing builder"),
170        ))
171        .with_delimiter(self.delimiter)
172        .with_batch_size(
173            self.batch_size
174                .expect("Batch size must be set before initializing builder"),
175        )
176        .with_header(self.has_header)
177        .with_quote(self.quote);
178        if let Some(terminator) = self.terminator {
179            builder = builder.with_terminator(terminator);
180        }
181        if let Some(proj) = &self.file_projection {
182            builder = builder.with_projection(proj.clone());
183        }
184        if let Some(escape) = self.escape {
185            builder = builder.with_escape(escape)
186        }
187        if let Some(comment) = self.comment {
188            builder = builder.with_comment(comment);
189        }
190
191        builder
192    }
193}
194
195/// A [`FileOpener`] that opens a CSV file and yields a [`FileOpenFuture`]
196pub struct CsvOpener {
197    config: Arc<CsvSource>,
198    file_compression_type: FileCompressionType,
199    object_store: Arc<dyn ObjectStore>,
200}
201
202impl CsvOpener {
203    /// Returns a [`CsvOpener`]
204    pub fn new(
205        config: Arc<CsvSource>,
206        file_compression_type: FileCompressionType,
207        object_store: Arc<dyn ObjectStore>,
208    ) -> Self {
209        Self {
210            config,
211            file_compression_type,
212            object_store,
213        }
214    }
215}
216
217impl From<CsvSource> for Arc<dyn FileSource> {
218    fn from(source: CsvSource) -> Self {
219        as_file_source(source)
220    }
221}
222
223impl FileSource for CsvSource {
224    fn create_file_opener(
225        &self,
226        object_store: Arc<dyn ObjectStore>,
227        base_config: &FileScanConfig,
228        _partition: usize,
229    ) -> Arc<dyn FileOpener> {
230        Arc::new(CsvOpener {
231            config: Arc::new(self.clone()),
232            file_compression_type: base_config.file_compression_type,
233            object_store,
234        })
235    }
236
237    fn as_any(&self) -> &dyn Any {
238        self
239    }
240
241    fn with_batch_size(&self, batch_size: usize) -> Arc<dyn FileSource> {
242        let mut conf = self.clone();
243        conf.batch_size = Some(batch_size);
244        Arc::new(conf)
245    }
246
247    fn with_schema(&self, schema: SchemaRef) -> Arc<dyn FileSource> {
248        let mut conf = self.clone();
249        conf.file_schema = Some(schema);
250        Arc::new(conf)
251    }
252
253    fn with_statistics(&self, statistics: Statistics) -> Arc<dyn FileSource> {
254        let mut conf = self.clone();
255        conf.projected_statistics = Some(statistics);
256        Arc::new(conf)
257    }
258
259    fn with_projection(&self, config: &FileScanConfig) -> Arc<dyn FileSource> {
260        let mut conf = self.clone();
261        conf.file_projection = config.file_column_projection_indices();
262        Arc::new(conf)
263    }
264
265    fn metrics(&self) -> &ExecutionPlanMetricsSet {
266        &self.metrics
267    }
268    fn statistics(&self) -> Result<Statistics> {
269        let statistics = &self.projected_statistics;
270        Ok(statistics
271            .clone()
272            .expect("projected_statistics must be set"))
273    }
274    fn file_type(&self) -> &str {
275        "csv"
276    }
277    fn fmt_extra(&self, t: DisplayFormatType, f: &mut fmt::Formatter) -> fmt::Result {
278        match t {
279            DisplayFormatType::Default | DisplayFormatType::Verbose => {
280                write!(f, ", has_header={}", self.has_header)
281            }
282            DisplayFormatType::TreeRender => Ok(()),
283        }
284    }
285
286    fn with_schema_adapter_factory(
287        &self,
288        schema_adapter_factory: Arc<dyn SchemaAdapterFactory>,
289    ) -> Result<Arc<dyn FileSource>> {
290        Ok(Arc::new(Self {
291            schema_adapter_factory: Some(schema_adapter_factory),
292            ..self.clone()
293        }))
294    }
295
296    fn schema_adapter_factory(&self) -> Option<Arc<dyn SchemaAdapterFactory>> {
297        self.schema_adapter_factory.clone()
298    }
299}
300
301impl FileOpener for CsvOpener {
302    /// Open a partitioned CSV file.
303    ///
304    /// If `file_meta.range` is `None`, the entire file is opened.
305    /// If `file_meta.range` is `Some(FileRange {start, end})`, this signifies that the partition
306    /// corresponds to the byte range [start, end) within the file.
307    ///
308    /// Note: `start` or `end` might be in the middle of some lines. In such cases, the following rules
309    /// are applied to determine which lines to read:
310    /// 1. The first line of the partition is the line in which the index of the first character >= `start`.
311    /// 2. The last line of the partition is the line in which the byte at position `end - 1` resides.
312    ///
313    /// Examples:
314    /// Consider the following partitions enclosed by braces `{}`:
315    ///
316    /// {A,1,2,3,4,5,6,7,8,9\n
317    ///  A,1,2,3,4,5,6,7,8,9\n}
318    ///  A,1,2,3,4,5,6,7,8,9\n
319    ///  The lines read would be: [0, 1]
320    ///
321    ///  A,{1,2,3,4,5,6,7,8,9\n
322    ///  A,1,2,3,4,5,6,7,8,9\n
323    ///  A},1,2,3,4,5,6,7,8,9\n
324    ///  The lines read would be: [1, 2]
325    fn open(&self, file_meta: FileMeta) -> Result<FileOpenFuture> {
326        // `self.config.has_header` controls whether to skip reading the 1st line header
327        // If the .csv file is read in parallel and this `CsvOpener` is only reading some middle
328        // partition, then don't skip first line
329        let mut csv_has_header = self.config.has_header;
330        if let Some(FileRange { start, .. }) = file_meta.range {
331            if start != 0 {
332                csv_has_header = false;
333            }
334        }
335
336        let config = CsvSource {
337            has_header: csv_has_header,
338            ..(*self.config).clone()
339        };
340
341        let file_compression_type = self.file_compression_type.to_owned();
342
343        if file_meta.range.is_some() {
344            assert!(
345                !file_compression_type.is_compressed(),
346                "Reading compressed .csv in parallel is not supported"
347            );
348        }
349
350        let store = Arc::clone(&self.object_store);
351        let terminator = self.config.terminator;
352
353        Ok(Box::pin(async move {
354            // Current partition contains bytes [start_byte, end_byte) (might contain incomplete lines at boundaries)
355
356            let calculated_range =
357                calculate_range(&file_meta, &store, terminator).await?;
358
359            let range = match calculated_range {
360                RangeCalculation::Range(None) => None,
361                RangeCalculation::Range(Some(range)) => Some(range.into()),
362                RangeCalculation::TerminateEarly => {
363                    return Ok(
364                        futures::stream::poll_fn(move |_| Poll::Ready(None)).boxed()
365                    )
366                }
367            };
368
369            let options = GetOptions {
370                range,
371                ..Default::default()
372            };
373
374            let result = store.get_opts(file_meta.location(), options).await?;
375
376            match result.payload {
377                #[cfg(not(target_arch = "wasm32"))]
378                GetResultPayload::File(mut file, _) => {
379                    let is_whole_file_scanned = file_meta.range.is_none();
380                    let decoder = if is_whole_file_scanned {
381                        // Don't seek if no range as breaks FIFO files
382                        file_compression_type.convert_read(file)?
383                    } else {
384                        file.seek(SeekFrom::Start(result.range.start as _))?;
385                        file_compression_type.convert_read(
386                            file.take((result.range.end - result.range.start) as u64),
387                        )?
388                    };
389
390                    Ok(futures::stream::iter(config.open(decoder)?).boxed())
391                }
392                GetResultPayload::Stream(s) => {
393                    let decoder = config.builder().build_decoder();
394                    let s = s.map_err(DataFusionError::from);
395                    let input = file_compression_type.convert_stream(s.boxed())?.fuse();
396
397                    Ok(deserialize_stream(
398                        input,
399                        DecoderDeserializer::new(CsvDecoder::new(decoder)),
400                    ))
401                }
402            }
403        }))
404    }
405}
406
407pub async fn plan_to_csv(
408    task_ctx: Arc<TaskContext>,
409    plan: Arc<dyn ExecutionPlan>,
410    path: impl AsRef<str>,
411) -> Result<()> {
412    let path = path.as_ref();
413    let parsed = ListingTableUrl::parse(path)?;
414    let object_store_url = parsed.object_store();
415    let store = task_ctx.runtime_env().object_store(&object_store_url)?;
416    let writer_buffer_size = task_ctx
417        .session_config()
418        .options()
419        .execution
420        .objectstore_writer_buffer_size;
421    let mut join_set = JoinSet::new();
422    for i in 0..plan.output_partitioning().partition_count() {
423        let storeref = Arc::clone(&store);
424        let plan: Arc<dyn ExecutionPlan> = Arc::clone(&plan);
425        let filename = format!("{}/part-{i}.csv", parsed.prefix());
426        let file = object_store::path::Path::parse(filename)?;
427
428        let mut stream = plan.execute(i, Arc::clone(&task_ctx))?;
429        join_set.spawn(async move {
430            let mut buf_writer =
431                BufWriter::with_capacity(storeref, file.clone(), writer_buffer_size);
432            let mut buffer = Vec::with_capacity(1024);
433            //only write headers on first iteration
434            let mut write_headers = true;
435            while let Some(batch) = stream.next().await.transpose()? {
436                let mut writer = csv::WriterBuilder::new()
437                    .with_header(write_headers)
438                    .build(buffer);
439                writer.write(&batch)?;
440                buffer = writer.into_inner();
441                buf_writer.write_all(&buffer).await?;
442                buffer.clear();
443                //prevent writing headers more than once
444                write_headers = false;
445            }
446            buf_writer.shutdown().await.map_err(DataFusionError::from)
447        });
448    }
449
450    while let Some(result) = join_set.join_next().await {
451        match result {
452            Ok(res) => res?, // propagate DataFusion error
453            Err(e) => {
454                if e.is_panic() {
455                    std::panic::resume_unwind(e.into_panic());
456                } else {
457                    unreachable!();
458                }
459            }
460        }
461    }
462
463    Ok(())
464}
datafusion_datasource_csv/source.rs

datafusion_datasource_csv/
source.rs
